Department of Physics

Physics of Galaxies
by
Peter Clegg
Revised by
P. Williams, B. Carr, D. Tsiklauri

Chapter 4. Active Galaxies

Physics of Galaxies Chapter 4

Table of Contents
CHAPTER 4: Active Galaxies ........................................................................................................................1
1. Introduction ........................................................................................................................................................... 1
2. Properties of Active Galaxies ................................................................................................................................ 1
2.1 Overview ......................................................................................................................................................... 1
2.2 Seyfert Galaxies............................................................................................................................................... 2
2.3 Quasars and BL Lacertae Objects.................................................................................................................... 3
2.4 Radio Galaxies................................................................................................................................................. 3
3. Models of Active Galactic Nuclei ......................................................................................................................... 4
3.1 The Source of Energy in AGN ........................................................................................................................ 4
3.1.1 Nuclear or Gravitational Power? .............................................................................................................. 4
3.1.2 Accretion Power ....................................................................................................................................... 4
3.1.3 The Eddington Limit................................................................................................................................. 5
3.2 Radiation Spectra............................................................................................................................................. 6
3.2.1 Characteristic Temperatures ..................................................................................................................... 6
3.2.2 The Radiated Spectrum of an Accretion Disc........................................................................................... 7
4. Models of Radio Sources....................................................................................................................................... 9
4.1 Synchrotron Radiation .................................................................................................................................... 9
4.1.1 The Frequency of Synchrotron Radiation................................................................................................. 9
4.1.2 The Power emitted by a Relativistic electron ........................................................................................... 9
4.1.3 The Synchrotron Spectrum ..................................................................................................................... 10
4.1.4 Lifetime of Relativistic Electrons ........................................................................................................... 10
4.2 Models of the Radio Lobes............................................................................................................................ 11
4.2.1 The Energy in the Lobes......................................................................................................................... 11
4.2.2 The Supply of Energy to the Lobes ........................................................................................................ 12
4.2.3 Motion of the Jets and Lobes.................................................................................................................. 13

Bibliography for CHAPTER 4 ................................................................................................................................ 15

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -i-

Physics of Galaxies
CHAPTER 4: ACTIVE GALAXIES
1. Introduction
The number of known types of galaxies showing
unusual activity has grown over the years to the point at
which the activity itself can hardly be considered
unusual, even if its origin is not well understood.
Moreover, similar sorts of activity, albeit on a smaller
scale, are seen in objects usually considered to be
“normal”. Many astronomers are therefore coming to
the view that we are seeing different mixtures, on
different scales, of one or two basic phenomena.
In this section, I shall first identify the various types of
active galaxy and discuss their properties. I shall then
review the whole phenomenon.
2. Properties of Active Galaxies
2.1 Overview
Active galaxies are frequently described as showing
violent or energetic activity. Perhaps a more useful
definition is galaxies displaying phenomena that cannot
be ascribed to normal stellar processes, although this
begs the question of what constitutes a “normal” stellar
process. It could be, for example, that the activity seen
in some nuclei is the result of a violent burst of star-
formation. In all cases, the activity seems to be
connected with the nucleus of the galaxy, whence the
term active galactic nucleus (AGN). In general, AGN
show some combination of the following features:
1. They have bright, star-like nucleus. In a short-
exposure image of the galaxy, only the point-like
nucleus is apparent. On longer exposure, the
nucleus saturates the detector – photographic plate
or CCD – and the rest of the galaxy becomes
apparent, although for quasars, this is only seen
with great difficulty.
2. The nucleus radiates over a very wide range of
wavelengths, from radio (in some cases) to X-rays.
This is unlike stars, which radiate mostly at optical
wavelengths, and normal galaxies, which are
dominated by starlight and the infrared emission
from cool interstellar dust.
3. The spectra of AGN are unlike those of normal
galaxies, being very blue and having no absorption
lines characteristic of starlight, but having strong
emission lines and covering a wide range of
wavelengths from the radio to X-rays and even
gamma rays.
4. The spectral energy distribution (SED) is the flux
density multiplied by the frequency1
SED := ν Fν , (2.1)
1
Another notation frequently used for the flux density is Sv.
Remember the units for flux density are Wm-2Hz-1.
© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010
Chapter 4
The SED of AGN are quite unlike those of normal
galaxies. For most AGN, the SED is roughly
constant but follows a power law from the hard X-
ray region of the spectrum to the far-infrared (far-
IR), the radiation being unpolarised. In radio-loud
AGN, this continues out to the radio region. In
blazars, the SED has a smooth broad hump and is
strongly polarised.
5. The SEDs typically have two “bumps” on top of the
power-law.
The blue bump is an excess, rising from the optical
to the ultraviolet (UV). There is some evidence that
this bump is seen descending again in the soft X-
ray region in which case it presumably peaks in the
extreme ultraviolet (XUV), at ν ~ 1016 Hz; this
corresponds to a temperature of ~ 300,000 K. The
peak cannot be observed directly because the
Galactic interstellar medium is opaque in this
region of the spectrum.
The X-ray bump occurs in the hard X-ray region
with Fν ∝ ν − (~ 0.7 ) [so that νFν is almost flat rising
only slowly with frequency] and is still rising at
E > 20 keV, corresponding to a temperature of
~ 108 K.
6. The strong emission lines are much too highly
excited to be the result of ionisation from OB stars
and are presumably ionised by the photons from the
blue bump. The lines can be very broad,
presumably the result of Doppler broadening with
velocities ~ 10,000 km s-1.
7. Many radio sources have double radio lobes,
symmetrical “blobs” on either side of the central
galaxy. These are very large structures, tens of
kiloparsecs in size and separated from the central
galaxy by ~ 50 kpc to 1 Mpc. The spectrum of radio
emission from these lobes is again a power law of
the form Fν ∝ ν − ( ~ 0.7 ) . There is often a bridge of
radio emission between the lobes and the central
galaxy.
8. Many radio galaxies have relativistic jets of
material pointing at the lobes. These jets seem to be
made up of blobs of material that appear to be
moving away from the galaxy faster than light!
Although we shall see that this is an optical
illusion, it does mean that the material in the jets
must be travelling at velocities close to that of light.
9. The brightness of AGN can vary very rapidly. The
X-rays from NGC 4051, for example, change by a
factor of 2 in 30 minutes! The variability tends to
depend upon wavelength:
X-ray Hours to days
Optical Weeks to years
Far-infrared Invariable
10. Since an object's output cannot change coherently
faster than the time it takes light to cross the
-1-
Physics of Galaxies
emitting region, this puts limits on the sizes of the
emitting regions (< 1 pc and for some even as small
as ~20 AU) and suggests enormous powers are
being generated in very small regions, sometimes
smaller than the solar system (see below).
11. The total power of AGN is hard to estimate because
of the great spread of wavelengths and also because
of the variability. The weakest known AGN are in
nearby galaxies with luminosities L ~ 1033 W,
whilst the most powerful distant quasars have
L ~ 1040 W. You should compare this with our
Galaxy, which has L ~ 1010 Lsun or ~ 1036 W. In the
most luminous cases, the nucleus can outshine the
whole of the rest of the galaxy!
Before trying to explain these observations, let us look
briefly at the various types of active galaxy; a fuller
discussion is given in [1]
2.2 Seyfert Galaxies
Carl Seyfert first noticed a class of spiral galaxy with
very bright nuclei in 1943. These were the first AGN to
be recognised and represent the mildest form of activity.
As well as dominating the light from the rest of the
galaxy, the output from Seyfert nuclei can vary in less
than a year. This sets an upper limit to the size of the
region responsible for the emission.
δt
∆t=∆r/c
∆r
Emission Detection at
event Earth
Figure 4-1. Limits on size of variable objects.
The upper part of Figure 4-1 shows schematically a
point source whose luminosity increases and decreases
on a time-scale δ t. The lower part of the figures shows
an extended source, each part of which brightens and
fades on the same time-scale δ t. The overall
brightening and fading of the source, as seen by an
observer, occurs over a longer time-scale ∆t because the
observed change in luminosity of the more distant parts
of the source arrives later than that of the nearer parts.
In general, we deduce that in an object whose
luminosity varies on a time-scale ∆t, the size of the
region responsible for the is emission cannot be larger
than ∆r, where
∆r <~ c∆t . (2.2)
© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010
Chapter 4
One light year, which is about a third of a parsec, is
extremely small compared with sizes of galaxies, which
are tens of kiloparsecs in diameter.
As with other AGN, the spectra of Seyferts differ from
those of “normal” galaxies. The latter have a continuum
spectrum – corresponding to the blackbody curves of
stars – containing the absorption lines of the stellar
atmospheres. Seyfert spectra have emission lines sitting
on top of their continua: this suggests the presence of
hot gas. The continua themselves, are not the thermal
spectra that are seen, for example, in HII regions.
Spectroscopy of Seyferts also distinguishes two types,
Seyfert 1 and Seyfert 2. In Seyfert 1 spectra, the
hydrogen Balmer lines (and other permitted lines) have
velocities some thousands of km s-1 wide. It is clear that
the thermal motions of individual atoms cannot produce
these widths: the temperatures necessary for such
thermal broadening would ionise the hydrogen and there
would no Balmer lines to be broadened! If the
broadening does not arise from the motion of the
individual molecules, it must be caused by bulk motion
of the gas: either the gas is rotating at high speed around
the centre of the galaxy or it is being expelled at high
speed. Remember that typical orbital velocities in
normal spiral galaxies are hundreds of km s-1. If the
widths are caused by rotation, it implies that the Seyfert
galaxies have masses of ~ 109 Msun within less than a
light-year of their centres. This represents a very high
density. Could Seyfert 1 galaxies harbour a black hole
in their centres?
Seyfert 1 spectra also contain forbidden lines which are
much narrower (hundreds of km s-1) than the hydrogen
lines. Presumably these lines are formed further from
the nucleus where the orbital velocities are lower. [The
velocities are still higher than typical in normal
galaxies.] Because they are forbidden lines, they must
certainly come from regions of relatively low density.
In Seyfert 2 spectra, permitted and forbidden lines have
the same widths, both being several hundred to a
thousand km s-1. It is likely that Seyfert 1s and 2s are
not completely different types of object, the broad-line
regions existing in Seyfert 2s but being obscured by
dust. There is support for this in the fact that Seyfert 2s
tend to be stronger infrared sources than Seyfert 1s,
suggesting that the ultraviolet radiation from the broad-
line region is absorbed by dust which then re-radiates it
in the infrared. Alternatively, Seyfert 2s could be
Seyfert 1s in which the central source of power has been
“turned off” so that the gas in the broad-line region is no
longer excited. This is a possible explanation for NGC
4151's having been seen to change from type 1 to type
2.
Seyfert galaxies make up about 1% of the population of
spirals, but 20-25% of the bright (L>1010Lsun) spirals.
An obvious interpretation is that 1% of spirals remain
Seyferts all their lives. There is an alternative
explanation, however. Because of the time-scales
involved, we are unable to wait long enough to see
galaxies evolve significantly. When we look at the sky,
therefore, we see a snapshot of all galaxies at one
-2-
Physics of Galaxies
moment in their evolution. It is possible that all galaxies
go through a “Seyfert phase”, spending 1% of their lives
as Seyferts. Since there are signs of some level of
activity in almost all galaxies, this is perhaps the more
likely explanation.
2.3 Quasars and BL Lacertae Objects
The first quasar to be identified was 3C48. As its name
implies, this is a radio source, being the 48th source in
the Third Cambridge Catalogue of radio sources. The
closest optical object to the radio position was a 16m
object whose image was indistinguishable from that of a
star, hence the name quasi-stellar (radio) source or
quasar. I have put the word radio in brackets because I
want to emphasise that, although the first quasars to be
identified were radio sources, we now know that only
about 1% or so of quasars are strong radio sources.
Radio-quiet quasars are often called QSOs; less often
(radio-loud) quasars are called QSRs. The terminology
is not universal or unambiguous.
Although 3C48 looked star-like on the photographic
plates, its spectrum was very different from that of a
star, being very blue and having unidentified broad
emission lines and a non-thermal continuum, very
similar to that of Seyfert 1s. The emission lines were
first identified by Maarten Schmidt of Palomar
Observatory. Working on another radio quasar, 3C 273,
he showed that the pattern of some lines could be
explained as hydrogen Balmer lines with what was, for
the time, a huge redshift of 0.158. Interpreting the
redshift as a Doppler effect gave a velocity of recession
nearly 16% of the velocity of light. If 3C273 were a
local object, its enormous velocity of recession would
be hard to explain. Nowadays, quasars with redshift
greater than 5 are known and it becomes almost
impossible to maintain that they are local objects. The
redshift of quasars is therefore generally accepted to be
cosmological in origin. The alternative explanation of
gravitational redshift is easily ruled out (cf. page 317 ff.
of [2]).
The problem with the cosmological interpretation of the
redshift of quasars is that it implies enormous
luminosities of up to about 1040 W. Moreover, like
Seyfert galaxies, the output of quasars can vary, on
time-scales of as little as days or even hours. A source
that can generate up to 1013 solar luminosities within a
volume comparable to that of the solar system is truly
awesome. Nevertheless, the consensus among
astronomers is that quasars are at cosmological
distances and that they are the active nuclei of galaxies2.
There are many arguments in support of this view. First,
the spectrum of a quasar can look so like the spectrum
of a Seyfert galaxy that it is tempting to believe we are
seeing extreme examples of Seyfert activity. To reverse
the argument, if bright Seyferts were moved out to the
distances of low-redshifts quasars, they would look like
2
This view is not universally held. Halton Arp, for example, argues
that some quasars at least are physically associated with galaxies that
have much lower redshifts. If this is the case, then some “new
physics” is required to explain the redshifts of these quasars.
© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010
Chapter 4
faint quasars. Secondly, the spectra of high-redshift
quasars exhibit a “forest” of Lyman alpha absorption
lines with a range of redshifts lower than that of the
quasar. These lines presumably arise in material, such as
the halos of faint galaxies lying between the quasar and
us. Lastly, modern techniques show the image – and the
spectrum – of the host galaxy around many quasars.
BL Lacertae objects (or BL Lacs) are named after the
prototype of these sources, the radio source BL
Lacertae. The brightness of these strange point-like
sources can vary by a factor of several on the time-scale
of days. In this respect, therefore, they are like an
extreme form of Seyfert nuclei or quasars. Most BL
Lacs, however, shows no emission lines in their spectra,
which are of typical non-thermal form. Their distances
are therefore difficult to estimate. Some BL Lacs have
faint evidence of material around the point source with
the spectral characteristics of a galaxy. This includes BL
Lac itself, whose redshift is 0.07, corresponding to a
distance of 200 h-1 Mpc.
2.4 Radio Galaxies
The optical luminosities of galaxies range from about
105 to 1010 Lsun. As might be expected from their ranges
of size and mass, ellipticals span this whole range whilst
spirals are confined roughly to the upper two decades.
Since the solar luminosity is 3.9 1026 W, the total optical
power radiated by galaxies ranges from 1032 to 1037 W.
Compared with these values, most galaxies radiate very
little power in the radio region of the spectrum,
luminous spirals emitting up to about 1033 W, for
example. There are, however, strong radio sources that,
in some cases, emit more radio power than the brightest
optical galaxies. Some of these radio sources are
quasars but not all quasars are radio sources.
Galaxy
Radio lobes
Figure 4-2. Cartoon of Cygnus A.
The strongest radio sources are typified by Cygnus A
(3C405), one of the most luminous. It has a radio output
of some 1038 W. Figure 4-2 is a schematic
representation of a low-resolution map of the radio
emission from Cygnus A. It shows the typical structure
of a “classical” double radio source, with two lobes of
radio material on either side of a peculiar galaxy and
about 50 kpc away from it. The core of the galaxy itself
is a radio emitter and I shall return to this shortly. Over
-3-
Physics of Galaxies
a wide range of frequencies, the flux density S(ν) from
the lobes has a power-law dependence on ν:
α
ν 
Fν = Fo ×   (2.3)
ν o 
or
log Fν = α logν + constant , (2.4)
where Fo is the flux-density at the frequency νo and α is
a constant called the spectral index3. Equation (2.4)
shows that the spectrum is a straight line on a log-log
plot of flux-density against frequency.
3. Models of Active Galactic Nuclei
3.1 The Source of Energy in AGN
3.1.1 NUCLEAR OR GRAVITATIONAL POWER?
We have to explain luminosities of at least 1039 W
(1013 Lsun), generated within regions only 1013 m or less
across and lasting for some 108 years. In other words,
we need to account for the generation of about 1054 J of
energy. [A related, but smaller, problem is the
explanation of the 1052 J or so found in the lobes of
radio galaxies.] Relativity tells us that, in order to
produce a luminosity L, we need to convert mass into
energy at the rate m& given by
dm L ; (3.1)
m& ≡ − =
dt c 2
the minus sign is convenient here because in order to
produce energy from mass we have to lose mass, so the
mass is decreasing with time. Equation (3.1) shows that
conversion of about a tenth of a solar mass per year
would produce 1039 W. At first sight, this seems modest:
a quasar would use up only 105 solar masses – perhaps
10-6 of the mass of a galaxy – in 108 years. What we
have not taken into account, however, is the efficiency
with which mass may be converted into energy. Let us
define the efficiency η of an energy-generating process
as the ratio of the energy E produced to the total rest-
mass energy of the material generating the power:
E kinetic energy of the body. We have to find a way of
η := . (3.2)
mc 2
Taking this efficiency into account, equation (3.1)
becomes
3
Note that some authors define the spectral index by:
−α
ν 
Fν = Fo ×   .
ν o 
This is because most of the early observations were of sources whose
spectral index was, by this definition, positive. It is becoming more
usual – but not universal – to use the form I have chosen.
© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010
Chapter 4
1 L. (3.3)
m& =
η c2
Suppose we try to produce 1039 W of luminosity by
nuclear burning as in stars4. The most efficient nuclear
process is the conversion of hydrogen into helium, with
η = 0.7%. Thus to generate an AGN luminosity of
1039 W by forming stars – that is the rate m& at which
interstellar material is converted into stars –which then
subsequently radiate by nuclear fusion while on the
main sequence is given by equation (3.3) as
10 39
m& ≈ kg s -1
(
0.007 × 3 × 108 ) 2 (3.4)
= 1.6 × 10 24 kg s -1 ≈ 25M sun y -1 .
This is a very high rate: in 108 y, a galaxy would
consume some 109 Msun, or perhaps 0.1% of its mass.
And, if we are to explain the variability of AGN, we
need all this mass to be contained within a region not
more than 1013 m across. The binding energy Ω of mass
m contained within a region R in radius is given by
Gm 2
−Ω~ (3.5)
R
and is about 1055 J for the above figures. This binding
energy, which could in principle be released by
gravitational contraction, is about an order of magnitude
more than we are trying to produce by nuclear means!
This suggests that the release of gravitational energy is
the more likely source of power in AGN, although we
need to look at the efficiency of this process as well.
3.1.2 ACCRETION POWER
Let us see if we can find a more efficient process.
Consider the decrease in gravitational potential energy
− ∆Ω of a body of mass m when it falls in from infinity
to within distance r of another body of mass M:
GMm
− ∆Ω = . (3.6)
r
If the body is in free-fall, however, this decrease in
potential energy will simply go into increasing the
converting the potential energy into radiation. The usual
model of this process is to consider the “body” to be
gaseous material in an accretion disc and which is
slowly spiralling into the central mass5. In order to
spiral, the material must be slowly losing angular
momentum and this is envisaged as arising from
viscosity in the disc – mutual friction between the
4
Even if we were to do this, we should still have to explain how to
convert ordinary stellar radiation into relativistic particles needed to
explain the observed spectra – see later.
5
This is almost certain to happen anyway because the in-falling
material will have angular momentum with respect to the central
object that it will have to get rid of.
-4-
Physics of Galaxies Chapter 4

material successive turns of the spiral6. This viscosity equation (3.9) and then only if the body of mass M is,
heats the disc, converting the kinetic energy of the indeed, a black hole. Because we get maximum
orbiting material into internal energy of the molecules efficiency from a black hole, let us assume that this is
and atoms of the disc. Finally, this heat is radiated away the case. In practice, therefore, the maximum radiant
by photons. Let us look at this process more energy Eradmax that I can extract from letting matter of
quantitatively. mass m fall into a black hole is given by
Since the material is in quasi-stationary orbit about the
max 1 GMm 1 GMm 1 (3.12)
central mass, the kinetic energy Ekin(r) and potential E rad = = = mc 2
energy Ω(r) of material at distance r from the centre 2 rlso 2 3rs 12
must obey the virial theorem:
with a corresponding maximum luminosity L given by
2 E kin (r ) + Ω(r ) = 0 . (3.7)
max
dErad 1
Lmax ≡ = m& c 2 . (3.13)
The total energy of the system – that is the sum of the dt 12
kinetic energy, the potential energy and any energy
Erad(r) that has been radiated away by the time the The minimum mass-inflow rate m& needed to produce a
material has reached r – must be conserved, so that luminosity L is therefore given by

Ekin (r ) + Ω(r ) + E rad (r ) = 0 , (3.8) m& = 12

Lmax (3.14)
c2
where I have taken the kinetic and potential energy of
the material at infinite distance from the central mass to Comparing equation (3.14) with equation (3.3), we see
be zero. Using equations (3.7) and (3.8) to solve for Ekin that the efficiency of gravitational accretion is 1/12 or
and Erad, about 8%, which is an order of magnitude more efficient
than nuclear processes. This means that only about
1 1 GMm 2 solar masses a year are needed to fuel the observed
E rad (r ) = − Ω(r ) = ; luminosities. Actually, we can expect the efficiency to
2 2 r (3.9)
be even higher than this. I have dealt only with a non-
1 1 GMm
E kin (r ) = − Ω(r ) = . rotating black hole. In practice, any black hole formed
2 2 r from the collapse of material to the centre of a galaxy is
likely to be rotating because the material from which it
Only half the potential energy is, therefore, available to formed would have had angular momentum about the
be converted into radiation. centre of the galaxy. The maximum efficiency of
It appears at first sight as if we could extract infinite extracting potential energy from a rotating black hole
energy from this process by allowing r to go to zero. can be shown to be about 40%.
There is, however, a natural limit to how close one can Note that the luminosity predicted by this model
approach to an object of mass M. General relativity tells depends only on the mass-infall rate and is independent
us that, if a body becomes smaller that its Schwarzschild of the mass of the black hole itself. The observed
radius rS, given by variability, however, limits the size of the region and
hence the Schwarzschild radius. Variability on the time
2GM scale of an hour demands a black hole mass of around
rS := , (3.10)
c2 108 - 109Msun, using equation (3.11).
3.1.3 THE EDDINGTON LIMIT
it will collapse to form a black hole from which nothing
can escape. Putting in numbers, we get We now show that there is a limiting “accretion-
luminosity” for a body of a given mass. Out-flowing
 M  photons exert a force on the in-flowing matter and, if
rS (light hours) = 2.76 × 10 −9  .
 (3.11) the flux of photons is large enough, this force will
 M sun  exceed the gravitational attraction of the central mass.
General relativity also says that there is no stable orbit Fout
closer to a black hole than 3rS. Material closer than this
plunges straight into the hole without having time electron
radiate away its remaining gravitational energy7. Hence proton
the radius rlso of the last stable orbit represents an r
absolute minimum to the value of r occurring in Fin

6
It has to be said that no satisfactory viscous mechanism has been
found!
7
This is a slight over-simplification but it serves our purpose. Figure 4-3. Origin of the Eddington Limit

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -5-

Physics of Galaxies Chapter 4

The details are as follows. The general relation between GM (me + mp ) GMmp
Fin (r ) = ≈ . (3.21)
the energy E and momentum ~ p of a relativistic particle r2 r2
of rest mass m is given by
Since the mass of the proton is some 2000 times that of
E =~
2
p 2c 2 + m 2c 4 . (3.15) the electron the approximation of neglecting me here is
an excellent one, so that the main inward gravitational
force is that on the proton
Since photons are massless, each carries momentum ~
p
given by GMm P
Fin (r ) = . (3.22)
~ r2
p=E (3.16)
c
To summarise: the radiation pressure acts on the
~ electrons which then communicate this to the protons
and the total force P carried per unit time by all the
photons emitted per second by a source of luminosity L via their electromagnetic coupling. The inward
is therefore given by gravitational force acts directly only on the more
massive protons, but again, this is communicated to the
~ L electrons via their electromagnetic interactions.
P= . (3.17)
c For accretion to take place, we need the inward force
given by equation (3.22) to be greater than the outward
and the total momentum per second per unit area carried force given by equation (3.20):
by these photons, the radiation pressure, is
GMmP Lσ T . (3.23)
 L/c . >
pγ (r ) =  r2 4π r 2 c
2
(3.18)
 4π r 
This means that, if a source is to derive its luminosity
Now the photons scatter off the hot, ionised in-falling from accretion without being blown apart by radiation
matter, mainly protons and electrons. The effective area pressure, its luminosity L must be less than the
presented by one electron or proton to the photon is the Eddington luminosity, LEddington:
Thomson cross-sections σT which, for a particle of
charge e and mass m, is GMmp c . (3.24)
L < LEddington := 4π
σT
2
2 e2 
σ T (e, m ) =   (3.19)
3  4πε o mc 2  Putting in numbers, we get

 M 
It is clear from equation (3.19) that, because protons are LEddington (W ) = 1.26 × 10 31  .
 (3.25)
some 2000 times more massive than electrons, the  M sun 
Thomson cross-section of electrons is some six orders
of magnitude greater than that of protons. The main Relation (3.24) can be re-arranged to give a lower limit
outward radiation pressure on the infalling matter is that on the mass of a source with a given observed
exerted by the photons scattering off the electrons. The luminosity L:
force Fout(r) exerted on one electron is therefore the
pressure times the Thomson cross-section σT of that 1 Lσ T . (3.26)
electron M > M min =
4π Gmp c
Lσ T
Fout (r ) = pγ (r ) × σ T = . (3.20) Again putting in numbers, we get
4π r 2 c
M min (M sun ) = 7.9 × 10 −32 L(W ) . (3.27)
Because the electrons in the plasma are coupled to the
protons through their interactions via electromagnetic Equation (3.27) shows that we need a black hole of at
forces, the outward force given by equation (3.20) is least 108 solar masses to give an accretion luminosity of
transferred to the protons as well. 1039 W.
In the overall neutral plasma of infalling matter there 3.2 Radiation Spectra
are equal numbers of protons and electrons. The
radiation pressure (3.20) acting outwards will then act 3.2.1 CHARACTERISTIC TEMPERATURES
against the gravitational force which acts inwards on Let us make some rough estimates of the temperatures
both electron and proton of total mass m=me + mp. The involved in accretion radiation. First, consider that
inward gravitational force Fin(r) corresponding to the coming from the inner edge of the accretion disc at
radiation pressure force (3.20) is therefore radius rmin. If we assume that the radiation is blackbody

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -6-

Physics of Galaxies Chapter 4

with effective temperature Teff, then the luminosity L is From equation (3.31), we find that the energy of the
given by photons corresponding to this temperature is about
100 MeV, many times the energy (1 MeV) required for
2
L ~ rminσ Teff4 , (3.28) the production of electron-positron pairs. From this, we
may tentatively conclude that:
where σ is the Stefan-Boltzmann constant. If we assume • the big blue bump arises from optically thick thermal
that the disc is radiating at the Eddington limit and that radiation from the inner parts of the accretion disc;
rmin=3rS is the radius of the last stable orbit, then we
have from equations (3.28), (3.24) and (3.10), • the hard X-rays and γ-rays originate from a region of
electron-positron plasma.
1/ 4
π mp c 5  3.2.2 THE RADIATED SPECTRUM OF AN ACCRETION DISC
Teff ~  . (3.29)
 9 GMσσ T  A complete derivation of the spectrum of the radiation
expected from an accretion disc is quite complicated
and is relegated to an appendix. I give a simplified
Putting in numbers, gives an estimate of the effective treatment here and merely quote the full result.
temperature of the accretion disc near the black hole,
Figure 4-1 shows schematically an accretion disc of
−1 / 4
7 M  density ρ(r) encircling a black hole and though which
Teff ~ 4 × 10  
 . (3.30) mass is spiralling down at a rate m& 9. The upper part of
 sun
M 
the figure shows the plan view and the lower part a
section.
For a 108 solar mass black hole, we find an effective
temperature of about 105 K. Using for the black body r + dr
peak frequency, r

hc
hν = ~ kT , (3.31)
λ
m&
we see that this radiation peaks at around 150 nm, or in
the UV to soft X-ray region of the spectrum where the
big blue bump lies. Moreover, from equation (3.10), we
see that the size of this region is about a light-hour
across, so we should not be surprised at variations on
the time-scale of hours.
h(r)
On the other hand, think of protons that fall straight
from “infinity” to rmin and are then thermalised. The
gain ∆Ekin in kinetic energy of each proton is given, Figure 4-4. Accretion disc.
from equation (3.9), by Since keeping track of signs consistently can cause
unnecessary confusion we shall drop all minus signs,
GMm p 1 keeping in mind that the inflow of mass at rate m& will
∆Ekin = = mp c 2 . (3.32)
3rS 6 cause a positive emission of radiation. Consider a thin
annulus of the disc of width dr at r. From section 3.1.2
If this energy is now thermalised to a temperature Ttherm, we see that rate of energy radiated by the material
we have falling in through this annulus is given by

3 1 & 1  GMm& 
∆E kin = kTtherm (3.33) dE& rad (r ) = − dΩ(r ) = d 
2 2 2  r  (3.36)
1 GMm&
so that = dr.
2 r2
2
1 mp c Since this is just the luminosity dL(r) of this part of the
Ttherm = (3.34)
9 k disc,

or, numerically 1 GMm&

dL(r ) ≡ dE& rad (r ) = dr . (3.37)
2 r2
Ttherm = 1.2 × 1012 K .8 (3.35)
9
If the disc is to be in a steady state, that is if there is to be no build up
of matter anywhere in it, the mass infall rate must be independent of
8
Note that these results are independent of the mass of the black hole. radius.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -7-

Physics of Galaxies Chapter 4

This power is radiated by an annulus of radius r and rmax

Lν = 2 × ∫ Fν (r ) × 2π rdr
thickness dr. In terms of the emerging flux F(r), the rlso
luminosity per unit area (Wm-2), of the disc at radius r rmax (3.45)
⌠  2π hν 3 1 
= 4π   2  rdr
dL( r ) = 2 × (2π rdr ) × F (r ) . (3.38) ⌡ c e hν kT (r )
− 1 
rlso
rmax
where the extra factor of two appears because the disc 8π 2 hν 3 ⌠ rdr
has two sides10. From equations (3.37) and (3.38), we =  hν kT (r )
c2 ⌡e −1
therefore have for the emergent flux, rlso

GMm& . where rmax is the outer radius of the disc and the first
F (r ) = (3.39) factor of two recognises that the disc has two faces.
8πr 3
Although the integration of equation (3.45) has to be
If the disc radiates as a black body, done numerically, we can get a good idea of what this
spectrum looks like by making some approximations. At
F (r ) = σ T 4 (r ) , (3.40) low frequencies, we have

where T(r) is the temperature of the disc at r and σ is the 1 kT hν

≈ ; << 1. (3.46)
Stefan-Boltzmann constant. From equations (3.39) and e hν kT
−1 hν kT
(3.40), we therefore get an estimate of the temperature
at r This is the Rayleigh-Jeans approximation. The lowest
temperature in the disc occurs at rmax. For frequencies
1/ 4
 GMm&  such that
T (r ) =   . (3.41)
 8π r 3σ 
hν << kT (rmax ) , (3.47)
A more detailed treatment gives
therefore, we have

 GMm& 
1/ 4
 r   1/ 2

T (r ) = 3 ×  × 1 −  lso   , (3.42) 8π 2ν 2 rmax

3  Lν ≈ k ∫ T ( r ) rdr
 8πr σ    r   c2 rlso
1/ 4 r
where rlso is the inner radius of the disc, assumed to be 8π 2ν 2  GMm&  max
= k  ∫ r
1/ 4
dr (3.48)
that of the last stable orbit. Except in the inner regions c2  8πσ  rlso
of the disc, equation (3.42) differs from (3.41) only by a 1/ 4
32π 2  GMm& k 4 5 
factor of three. I shall continue to use the simple form. ≈  rmax  ν 2
5  8πσc 8 
The specific intensity Iν(T) (units Wm-2Hz-1ster-1) of
black-body radiation at temperature T is given by ∝ν 2,
Planck’s formula
where, in the third line, I have assumed that rlso << rmax.
2 hν 3 1 At low frequencies, therefore, the spectrum is a power
Iν (T ) = 2 (3.43) law.
c e hν kT − 1
At high frequencies, we have
so that the flux density Fν(r) (units Wm-2Hz-1) at the
accretion disc is given by11 1 hν
hν kT
≈ e − hν kT
; >> 1. (3.49)
3
e −1 kT
2π hν 1
Fν (ν , r ) = , (3.44)
c 2 e hν kT (r ) − 1 The highest temperature in the disc occurs at rlso. For
frequencies such that
where T(r) is given by equation (3.41). The total
luminosity density Lν (units Wm-2Hz-1) of the disc is hν >> kT (rlso ) , (3.50)
therefore given by
therefore, we have

10
I am assuming that the disc is optically thick.
11
Note that a black surface radiates into π steradians.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -8-

Physics of Galaxies Chapter 4

8π 2 hν 3
rmax the electron will emit electromagnetic radiation with
∫e
− hv kT ( r )
Lν ≈ rdr frequency νc. This is cyclotron radiation. If the electron
c2 rlso is moving relativistically with velocity u however, it
r emits radiation over a broad band of frequencies peaked
8π 2 hν 3 − hv kT (rlso ) max ,
≈
c2
e ∫ rdr
rlso
(3.51) at frequency ν given by

γ2 1 eB 2
4π 2 hν 3 2 − hv kT (rlso ) ν= νc = γ , (4.2)
≈ rmax e 2 4π m e
c2
where the Lorentz factor γ is given by
where, in the second approximation, I have replaced the
exponential term with its maximum value and, in the
1 1
third, have again assumed that rmax >> rlso. At high γ = = ;
2 2
frequencies, the spectrum falls off exponentially with 1− u c 1− β 2 (4.3)
frequency. u
β := .
What about in between? There is no simple argument I c
can give here, but we expect a power law for Lν
somewhere between ν0 and ν1/3 in the intermediate Using the fact that the total energy E of the relativistic
region. The overall spectrum is sketched in Figure 4-5. electron is given by
This is not bad a representation of the “blue bump” in
AGN spectra. E = γ me c 2 , (4.4)

we get
ln Lν

∼ν0 to ν1/3 1  e  2
-hν/kT ν=   BE = aBE 2 , (4.5)
e
4π  m 3c 4 
 e 

ν2 where

1  e 
a :=  . (4.6)
4π  m 3c 4 
 e 
ln ν
Putting in numerical values, we get
Figure 4-5. Overall spectrum of accretion disc.
ν (Hz ) = 2.1 × 10 36 B(T )E 2 (J ) . (4.7)

4. Models of Radio Sources

4.1.2 THE POWER EMITTED BY A RELATIVISTIC
4.1 Synchrotron Radiation ELECTRON
4.1.1 THE FREQUENCY OF SYNCHROTRON RADIATION The power P(E) radiated by an electron of energy E
A power-law spectrum of the form given in equation spiralling in a magnetic field is given by
(2.3) is consistent with the synchrotron radiation
emitted by a collection of charged particles, of various 2 1  e4 B 2 
P (E ) =  (βγ )2
energies, moving with relativistic velocities in a 3 4πε o  m 2c 
magnetic field. Synchrotron emission is also consistent  e 
(4.8)
with the observed polarisation of the radiation. I shall 1  e 4 
 × β 2B2E2
=
give a simplified treatment of such radiation. 6πε o  me4 c 5 

An electron moving perpendicularly to a magnetic field
of induction B (units Tesla, abbreviation T) describes a where εo is the permittivity of free space and where the
circular orbit with the cyclotron frequency νc given by second line come from equation (4.4).
1 eB In any collection of electrons in space, there must be an
νc = , (4.1) equal number of protons to preserve charge neutrality,
2π me
and these protons will also radiate in a magnetic field.
Equation (4.8) shows, however, that the power radiated
where me is the rest-mass of the electron. Notice that νc by a particle in inversely proportional to the fourth
is independent of the velocity (or energy) of the power of its mass. Even if the protons are moving
electron. Because circular motion involves acceleration, relativistically with the same energy as the electrons,

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 -9-

Physics of Galaxies Chapter 4

therefore, the power they radiate will be relatively where I have used equation (4.10). We can use equation
negligible because they are some two thousand times (4.5) to eliminate the energy E in favour of the
more massive12. frequency ν , ν ∝ BE 2 :
In all the cases we shall deal with, the electrons are
(2 − p ) / 2 1/ 2
ultra-relativistic so that ν  ν 
dL(ν ) = N o E op bB 2   d 
β ≈1. (4.9)  aB   aB 
 1+ p   1− p 
(4.15)
1  b     
We can therefore re-write equation (4.8) as = N o E op  (3− p ) / 2  B  2   2 
ν dν .
2 a 
P(E ) ≈ bB 2 E 2 , (4.10)
This spectrum has the same form as that given in
equation (2.3) provided that
where
1− p
1  e 4 
.
α=
2
. (4.16)
b= (4.11)
6πε o  me4 c 5 

Using the local cosmic ray energy spectrum of the form
Putting in numerical values, we get (4.13) with p ~ 2.5, gives α ~ −0.75. This is remarkably
close to the commonly observed value of −0.7. We may
P(W ) = 2.4 × 1012 B 2 (T )E 2 (J ) . (4.12) therefore feel that we have “explained” the power law
spectrum of AGN in terms of “cosmic rays” in the radio
4.1.3 THE SYNCHROTRON SPECTRUM sources.

To find the spectrum of synchrotron radiation emitted

by the electrons we need to know the energy spectrum
of the electrons. We have a clue from the spectrum of
cosmic ray electrons observed on the earth: they have a N ( E ) ∝ E −2.5
power law spectrum, Figure 4-6. Moreover there is a
theory for how cosmic rays are accelerated which
explains this power law spectrum: the electrons are
accelerated when they pass through shocks in supernova
explosions. It is clear from examining the images of
extragalactic jets that there are also shocks present there
and there is every reason to believe shocks occur closer
in to the accretion disc around the black hole. This leads
us to suppose in our population of radiating electrons
the number N(E)dE of relativistic electrons with
energies between E and E + dE is given by a power law

  E − p
 N   dE , Emin < E ≤ Emax
N (E )dE =  o  Eo  (4.13)

0 otherwise.

with p ~ 2.5 being typical of the energy spectrum of

cosmic ray electrons observed at the earth, Figure 4-6.
In a magnetic field each electron of energy E radiates
power P(E) given by (4.10) so that the luminosity dL(E)
of the electrons with energies in the range E to E + dE is
given by

dL(E ) = N (E ) dE × P (E )
(4.14) Figure 4-6. Energy spectrum of cosmic ray electrons
= ( N o E op bB 2 ) E 2 − p dE , arriving at the top of the earth’s atmosphere.

4.1.4 LIFETIME OF RELATIVISTIC ELECTRONS

Because the electron is radiating power P(E), it is losing
energy at this rate so
12
Equation (3.41) shows that they will also radiate at much lower
frequencies.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 10 -

Physics of Galaxies Chapter 4

dE 4.2 Models of the Radio Lobes

− = P (E ) . (4.17)
dt

Substituting for P(E) from equation (4.10) we find 4.2.1 THE ENERGY IN THE LOBES
In this section, I shall assume that all electrons have the
dE same energy Eo. That is, I shall replace the true energy
− = bB 2 E 2 . (4.18)
dt spectrum given by equation (4.13) by an extremely
simple spectrum where the population consists of a total
Very approximately by replacing infinitesimals by finite of N0 electrons all with the same energy E0. This will
quantities, we estimate the time ∆t for an electron with make the discussion much easier without changing the
energy E to lose energy ∆E overall conclusions. Since now all N0 electrons radiate
the same power P(E0) given by (4.10) the total
∆E luminosity is then simply,
∆t = . (4.19)
bB 2 E 2
L = N o P (E o ) , (4.26)
Even more cavalierly, we put ∆E equal to E and obtain a
characteristic time τ(E) for an electron to lose its and the total energy Eelectrons contained in the electrons is
energy, where given by

1 E electrons = N o E o . (4.27)
τ (E ) = 2
. (4.20)
bB E
From equations (4.8) and (4.5) we get
Notice that the more energetic the electron, the more
N o Eo Eo
rapidly it loses its energy. τ(E) can be considered as the Eelectrons = L= L
lifetime for radiation from an electron of energy E. N o P (Eo ) P (Eo )
(4.28)
More formally, we can integrate equation (4.18):  a1 / 2  − 3 / 2 −1 / 2
=   LB ν o ,

E  b 
⌠ dE ′
 2 = −bB 2 t , (4.21)
⌡ E′ where I have used equation (4.5) to eliminate Eo. Putting
E0 in numerical values, we obtain

Eelectrons (J ) = 6 ×10 5 L(W )B(T ) ν o (Hz )−1/ 2 . (4.29)

−3 / 2
where E0 is the initial energy of the electron at t = 0.
Carrying out the integration on the left-hand side, we
find the energy after time t is Equation (4.28) takes account only of the energy in the
electrons. If the material is to be electrically neutral, the
E0 . (4.22) electrons must be accompanied by an equal number of
E (t ) =
1 + bB 2 E0t protons, which will also have energy. To allow for this,
let us say that the energy Eprotons in the protons is a fixed
The time taken for the energy to decay to half its initial multiple K of that in the electrons, Eelectrons
value, E=E0/2 is given by
E protons = KE electrons . (4.30)
1
τ (E ) = 2 . (4.23)]
bB E The total energy Eparticles in particles is then given by

where I have, after the calculation, replaced the symbol E particles ≡ E electrons + E protons
E0 by E. This confirms our simple-minded guess (4.20).
= (1 + K )E electrons (4.31)
We can use equation (4.5) to eliminate the energy E
 a1/ 2  −3 / 2 −1 / 2
from (4.20) and obtain the characteristic lifetime τ(ν) of = (1 + K )  LB νo .
 b 
an electron radiating at frequency ν:  

 a 1 / 2  −3 / 2 −1 / 2 We have no way of determining K directly. We may

τ (ν ) =  B
 ν . (4.24) have a clue in the cosmic ray flux measured near the
 b  Earth in which the protons carry 100 times as much
energy as the electrons. In the absence of better
Substituting numerical values, we get information, therefore, let us take K to be 100. Then for
example, the total particle energy in the radio lobes of
τ (sec ) = 6 × 105 B(T )−3 / 2ν (Hz )−1 / 2 . (4.25) Cygnus A is equivalent to the rest-mass of about 105
Msun.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 11 -

Physics of Galaxies Chapter 4

A magnetic field of induction B has an energy density To find the minimum, we differentiate Etotal with respect
ufield associated with it, given by to B and set the result equal to zero to get Bmin. We find

 3 a1/ 2   Lν −1/ 2 

2/7
B2
u field = , (4.32) Bmin =  µo (1+ K ) o  . (4.36)
2µ o  2 b   V 
where µo is the permeability of free space. The total The first term in round brackets depends solely on
energy Efield contained in the field is therefore given by fundamental constants. The terms in square brackets are
estimated or observed quantities.
B2
E field = V , (4.33) Putting in numerical values, we get
2µ o
2/7
 L(W )ν o (Hz ) 
−1 / 2
. (4.37)
where V is the volume of the emitting region. Bmin (T ) = (1 + K ) 
Numerically, we have  V m 3
( ) 

Efield (J ) = 2.3 × 1064 V (kpc) B(T ) 2 ,

3
(4.34) Equations (4.31) and (4.34) then show that this
magnetic induction Bmin makes the energy in the
The total energy Etotal in the lobes needed to give the particles nearly equal to that in the magnetic field:
observed luminosity is therefore given, from equations
(4.28) and (4.33), by E particles 4
= . (4.38)
E field 3
E total ≡ E particles + Efield
 a1 / 2  −3 / 2 −1 / 2 Some people prefer to start with the assumption that the
B2 (4.35)
= (1 + K )  LB ν +V . energy is equally divided between the particles and the
 b  o
2µ o
  field – the so-called equipartition of energy – rather than
to assume that the energy is minimised. From equation
Unfortunately it is rarely possible to measure the (4.38), it is obviously immaterial which assumption is
magnetic field independently. How, therefore, are we to made.
estimate the total energy Etotal? The first term in From equations (4.33), (4.37) and (4.38), we have for
equation (4.35) is a decreasing function of the magnetic the minimum total energy Emin,
induction B whereas the second term is an increasing
function of B. There must therefore be a minimum in
 4
the total energy needed to give an observed luminosity E min = 1 +  E field,min =
L. This is shown in Figure 4-7, which plots the  3
logarithm of the particle, field and total energies against 3/ 7 4/7 (4.39)
7 V  3 a1/ 2 
the logarithm of the induction. =    (1 + K ) Lν o−1 / 2  .
6  µo   2 b 

log E 4.2.2 THE SUPPLY OF ENERGY TO THE LOBES

Perhaps the most natural assumption, given their

Etotal
Efield Eparticles
log B
Figure 4-7. Variation of particle (dashed curve), field
(dotted curve) and total (solid curve)energies with B.
It is usually assumed that the magnetic field has the
value that minimises the total energy for no better
reason than that, even with this assumption, the total
energies required – of around 1053 J – are
embarrassingly high!
disposition on either side of the galaxy, is that the radio
lobes were ejected as entities from the galaxy and have
since been travelling independently through
intergalactic space. There is certainly nothing in the
vicinity of the lobes themselves that could be
responsible for them. The lobes are tens of kiloparsecs
or more away from the galaxy – in the case of Cygnus A
some 50 kpc or 150 000 light years. As we shall see
later, they are moving out into the intergalactic medium
at no more than a tenth of the velocity of light so that
they would have taken ~106 years to get there. Yet it can
be shown from equation (4.25) that the lifetime of an
electron in a radio lobe is typically ~105 years. The
electrons would therefore have lost all their energy in
the time taken by the lobes to get to their present
position.
In fact, the problem is far worse than appears at first
sight! If the lobes had been ejected from the central

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 12 -

Physics of Galaxies Chapter 4

galaxy, it is reasonable to assume that they had come For changes that are slow compared with the time to
from the nucleus of the galaxy – where the activity is complete one orbit – that is, for adiabatic changes – Φa
seen – rather than from the benign elliptical galaxy must be constant so that
surrounding the core. But the variability – on the time-
scale of a year or less – of the activity in the nucleus E ′2 E 2
shows that it must be less that a few parsecs in size. Let = (4.46)
us take an upper limit of 10 parsecs. The radio lobes B′ B
themselves, however, are of the order of several
or, using equation (4.18),
kiloparsecs. Hence, the lobes must have undergone
expansion by a factor of around 1000 in their passage −1
from the nucleus of the galaxy. Let us denote the factor E′ = f E. (4.47)
of expansion by f. Then we can say that the lobe
expands form an initial size d to a final size d' given by If no new electrons are added to the plasma, the total
fd. Symbolically: number N of electrons contained in the lobes must be
constant so that
d → d ′ = fd . (4.40)
′
E particles = (1 + K )NE ′
(4.48)
Now the lobes contain plasma – highly ionised matter – = (1 + K )Nf −1
E= f −1
E particles .
and the magnetic field B is therefore “frozen” into the
matter13. As the lobes expand, therefore, the field lines
The field and particle energies therefore scale in the
get further apart and the field itself decreases.
same way with the expansion factor f, preserving the
Quantitatively, the flux threading the lobe must be
“equipartition” between the particle and field energies:
conserved. Since the size of the lobe is d, the flux Φ
threading it is given by Using equation (4.10) for the power radiated by an
individual electron, we have for the total power P′total
Φ ~ Bd 2 = B ′d ′ 2 = constant , (4.41) radiated by the lobes after expansion

where B′ is the final value of the induction. From ′ = NbB ′ 2 E ′ 2

Ptotal
equations (4.40) and (4.41), therefore, (4.49)
−4 −2 −6
= Nbf B2 f E2 = f Ptotal ,
−2
B′ = f B. (4.42)
where Ptotal is the total power radiated before the
expansion. Equation (4.49) says that the power radiated
The total energy E′field within the field after expansion is by the lobes decreases by a factor f 6 during the
given by expansion. As we have seen, f is of order 103 so the
power decreases by eighteen orders of magnitude during
B′ 2 B′2 the expansion! If the ejection hypothesis were true,
′ =V ′
E field ~ d ′3
2µ o 2µ o therefore, one would expect to find that sources with
(4.43) small separation between the lobes and the central
−4
3 3 f B2 −1 galaxy were on average more powerful than the larger
= f d = f E field ,
2µ o sources. On the contrary, it is the larger sources that
are, on the whole, more powerful. We must therefore
where V' is the expanded volume of the lobes and Efield seek another explanation of the origin of the radio lobes.
is the total field energy before expansion.
The accepted explanation is that there is a beam or jet of
What about the electrons? The radius a of the orbit of a particles emanating from the central galaxy which feeds
highly-relativistic electron of energy E spiralling around the lobes with fresh relativistic electrons. As we have
the magnetic field B is given by seen, there is direct evidence for such beams in radio
sources, as well as in their synchrotron power law
E spectrum.
a= (4.44)
eBc 4.2.3 MOTION OF THE JETS AND LOBES
so that the flux Φa threading the electron's orbit is given 4.2.3.1 Superluminal Velocity in Jets
by
Figure 4-8 shows knots or blobs of material in a jet
2 leaving a central galaxy with velocity V. By taking radio
1E
Φa ~ a2B = 2 2
. (4.45) measurements – separated by months or years – of these
e c B blobs, we can measure their proper motion µ, that is
their angular velocity in the plane of the sky. If we can
establish the distance ro of the galaxy, we can calculate
the rate of change p& of the projected distance p, which is
13
Strictly, the conductivity of the plasma would have to be infinite for
the field to be completely tied to the matter, but the approximation is
its apparent velocity:
good.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 13 -

Physics of Galaxies Chapter 4

p& = ro µ . (4.50)  ∆p 
v 
p& ≡
∆p
=
∆p
=  ∆l  . (4.54)
∆r
∆t ∆l − ∆r v  ∆r 
galaxy D ro
1−  
A θ v c c  ∆l 

In the last step I have multiplied numerator and

∆p denominator by v/∆l to obtain a form in which the
∆l bracketed ratios can now easily be expressed in terms of
the angle θ in the right-angled triangle ABD in Figure 4-
r 8 with sides ∆p, ∆r and ∆l:
B
∆p ∆r
= sin θ and = cosθ . (4.55)
V ∆l ∆l
Figure 4-8. Illustration of superluminal motion. Finally, dividing equation by c and defining the
observed and actual beta factors as
As we have already seen, p& is found to be greater than
the velocity of light c in several sources. I shall now p& v
β obs := and β := (4.56)
show that this is an optical illusion and does not imply c c
that the blobs are violating the tenets of relativity.
Suppose the furthest blob marked B was first observed
being ejected from the galaxy at A and then observed a
β sin θ
β obs = , (4.57)
time ∆t later to have moved to the position shown in 1 − β cosθ
Figure 4-8. The radiation emitted by the blob on
emission from the galaxy took a time t0 to reach the This is our final result. This formula is plotted in Figure
earth 4-9 and Figure 4-10 where we see that superluminal
apparent velocities βobs = p& / c even as high as 10 can be
ro accounted for provided the jet emerges at a small angle
t0 = , (4.51)
c and a large but subluminal speed. To observe a blob
apparently moving faster than, or at the speed of light
where ro is the distance of the galaxy from Earth14. The requires
blob then travels a distance ∆l at speed v to reach the
position B depicted in Figure 4-8, after which the  β sinθ 
radiation we observe from the blob travels a distance r
β obs ≥ 1, i.e.   ≥ 1. (4.58)
1− β cosθ 
to the earth; this amounts to a total time
We can re-arrange equation (4.58) to give
∆l r
t= + . (4.52)
v c 1
β≥ (4.59)
sin θ + cosθ
Equations (4.51) and (4.52) then give the total time
∆t =t - t0 earth observers see the blob take to move out The minimum value βmin of β needed for apparent
from the galaxy at A to the position B in Figure 4-8 superluminal velocity is 1/√2 corresponding to θ = π/4:

 ∆l r  r ∆l ∆r c
∆t = t − t 0 =  +  − 0 = − (4.53) Vmin ≡ cβ min = = 0.707c . (4.60)
 v c c v c 2

where ∆r = r0 – r is depicted in Figure 4-8. Note that the
time difference ∆t between the reception of the two
signals is less than the time between emission ∆l/v
because the light emitted from B has less far to travel: it
is this difference which gives rise to the optical illusion.
In the time ∆t the blob is observed to have moved a
distance ∆p in the plane of the sky (perpendicular to the
line-of-sight) and therefore to have apparently travelled
at a speed,
This is illustrated in Figure 4-9. The fact that we do see
such motion shows that the material in the jets feeding
the radio lobes is moving relativistically15. Note that
(sinθ + cosθ) is symmetrical about θ = π/4 so that the
same value of β is needed for the angle (π/4 − ∆θ) as for
(π/4 + ∆θ). The reason for this is that, what we lose in a
longer light travel time (cosθ decreasing) we gain in

14
For simplicity, I assume that the galaxy is at rest with respect to the
Earth. This assumption does not affect the conclusion.
15
It is important to realise that we are here speaking of the bulk
motion of the jets. The electrons within the jets may also moving with
different relativistic velocities, e.g. spiraling around the magnetic field
lines.

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 14 -

Physics of Galaxies Chapter 4

greater projected distance (sinθ increasing), and vice

versa.
Inequality (4.59) can be manipulated to give the range
of values of θ between which we can get superluminal
velocities for a given value of β:

 1   −1  1  
sin −1  2
1 − 2 β − 1  ≤ θ ≤ sin 
2
1 + 2 β − 1 
 2β    2β  

(4.61)

1.5

1.25

1
βobs ( 0.6 , θ )

βobs ( 0.707, θ )
0.75
βobs ( 0.8 , θ )

c(θ)
0.5

0.25

0
0 10 20 30 40 50 60 70 80 90
θ

Figure 4-9. Observed jet velocities for actual velocities

β=0.6(dotted), 0.707(solid), 0.8 (dashed).

10

7
βobs ( 0.95, θ )
6
βobs ( 0.990, θ )
5
βobs ( 0.995, θ )
4
c(θ)
3

0
0 18 36 54 72 90
θ

Figure 4-10. Observed apparent jet velocities for

actual velocities β=0.95(dotted), 0.99(solid), 0.995
(dashed).

In Figure 4-9 and Figure 4-10 the two ends of this range
in θ correspond to the points at which the curves cross
the line βobs=1; as β approaches 1 notice how these
values approach θ = 0 and 90°, Figure 4-10.

Bibliography for CHAPTER 4

[1] Robson, I. Active Galactic Nuclei, Wiley 1996.

ISBN 0-471-96050-0
[2] Shu, F H. The Physical Universe, University
Science Books, CA, 1982. ISBN 0-19-855-706-X

© P E Clegg 2001, revised B J Carr 2007, D Tsiklauri 2010 - 15 -